The average starting salary for students graduating with a bachelor's degree in computer engineering (2016-17).
Aerial view of the 400-ft telescope. The full-size 1957 sedan parked at the far end of the reflector gives scale.
by George W. Swenson, Jr.
In 1956 the University of Illinois invited the writer to join its faculty to design and build a large radio telescope. Professor George C. McVittie, Head of the Astronomy Department, believed with many other cosmologists that a very extensive and complete catalog of discrete, cosmic radio sources would help to distinguish among competing cosmological theories. Two major catalogues of sources had been published by radio astronomers in Cambridge, England, and Sydney, Australia, but they did not agree well in the region of the sky in which they overlapped and it was desirable to confirm them with different instruments. McVittie and Professor Edward C. Jordan, Head of the Electrical Engineering Department, agreed that such a program would be an appropriate area for cooperation between the departments and sought an engineer to undertake the job.
Copyright (c) 1986 Institute of Electrical and Electronics Engineers. Reprinted from the IEEE Antennas and Propagation Society Newsletter, vol. 28, no. 6, pp.13-16, December 1986.
This material is posted here with permission of IEEE. Internal of personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution must be obtained from the IEEE by sending a blank e-mail message to firstname.lastname@example.org.
The first task was to study the existing instruments engaged in cosmic radio source cataloguing, to determine the errors to which they were most susceptible in order to design a new instrument which would best complement them. It was also desired to compile a deeper catalog than existed; that is, a catalog complete to a lower flux level.
The University in 1957 sent me on a tour of the world's most prominent radio observatories, including those at Cambridge and Manchester (Jodrell Bank), England; Sydney, Australia; Paris, Nancáy, and Haute Provence, France; and Dwingeloo, Netherlands. At that time the best-known catalogs had been compiled by Martin Ryle's group at Cambridge and Bernard Mills' group at Sydney. The Cambridge instrument was a compound interferometer operating at 159 MHz. One of its problems involved misidentification of the reception lobes of the interferometer, thus causing ambiguities in source position. The Sydney instrument, the Mills Cross, operated at 81.5 MHz and involved the multiplication of orthogonal fan beams produced by two linear arrays, in order to produce a narrow pencil beam. As the sidelobes of one array were multiplied by the main beam of the other, the resulting sidelobes of the instrument were inherently high. A strong source in a sidelobe could easily be taken for a weak source in the main beam.
It was apparent that a new instrument design should emphasize sidelobe reduction, should involve a different frequency range from existing instruments, and should be large enough and have sensitive-enough receivers to permit detection of sources an order of magnitude fainter than those in existing catalogs. The weaker the source to be catalogued the narrower must be the main beam and the lower the sidelobe levels, in order to avoid the "confusion" problem that had plagued earlier efforts. I thought a filled-aperture antenna had the best chance of achieving these goals; however, it must be very large by the standards of the day. How to achieve the best compromise among size, surface precision, frequency, and cost was the problem.
Having learned the rudiments of the cosmic cataloguing business from Bernard Mills and Joseph Pawsey at Sydney, I mulled these questions over while touring the world's radio telescopes. At the Observatoire de Haute Provence, Dr. M. Laffineur had built an interferometer, each of whose elements was a 100-foot square (approximately) parabolic cylindrical reflector of poultry wire whose cylindrical axis was north-south. It was fed by a linear array of dipoles along the focal line, and the whole system was suspended above ground on wooden poles. Phase adjustment of the dipoles permitted steering of the main beam in declination, while the earth's rotation scanned the beam in right ascension. This "meridian transit" system is acceptable for a systematic survey and does not require any mechanical motions of the structure. It seemed a very economical scheme. I discussed it with Robert Hanbury Brown at Jodrell Bank and we agreed it might be a good approach to the problem.
There was a difficulty involving the precise phasing of the dipoles along the focal line. At the Carnegie Institution of Washington John Firor had built an array for solar studies in which he used helical antennas mechanically rotated about their helical axes to adjust their relative phases. Each element in Firor's array consisted of two helices of opposite sense, combined to be sensitive to linear polarization. If circular polarization were acceptable, each element might be a single helix; however, the off-axis radiation is elliptical and this complicates the problem of appropriately illuminating a reflector antenna.
With these thoughts in mind, the conceptual design of the radio telescope was begun. Several design parameters were quickly obtained. The conventional wisdom of the day held that a resolution of several score beamwidths per source was necessary to avoid confusion. Adopting a lower flux level of 10-26 W/M2/Hz and extrapolating on the then-known source number/strength statistics led to a required beam-width of about 0.25 degree. A frequency of 600 MHz would be sufficiently higher than that of existing catalogs to yield useful spectral data and was still in the range where sensitive receivers were possible. Thus, the linear dimension of the antenna should be about 120 meters, and the configuration should be roughly square or circular. A reasonable focal-length-to-diameter ratio to permit good illumination of the reflector is 0.4. This would require very high structures if constructed on level ground, especially expensive if strong enough to withstand winter loading by snow and ice, and if precise enough to operate efficiently at l = 50 cm. A suspended structure was briefly considered but the construction costs and the effort needed to maintain dimensional precision promised to be excessive.
The alternative was to build the reflector of earth, in large part below ground level. An extensive search disclosed a stream valley near Danville, Illinois, whose axis was nearly north-south and which had appropriate dimensions. The land was available for purchase. The geological conditions were acceptable from the structural viewpoint. The natural dimensions of the valley dictated only minor changes in the nominal parameters; the final values are: width, 400 feet; f/d, 0.39; length of reflector surface, 600 feet; height of towers, 165 feet.
Television channel No. 37, 608-614 MHz, was at that time unassigned and unused in North America and the corresponding channel was unassigned elsewhere in the world. After a lengthy legal and political struggle, this frequency band was assigned to radio astronomy on a worldwide basis. This outcome was years in the making and represents a story all by itself. In the meantime, the telescope was completed, relying on hope and faith that the frequency use question would somehow be resolved in a satisfactory way.
The basic parameters and construction technique having thus been established, the multifarious details of the design remained to be determined. For help in this we turned to the outstanding Antenna Laboratory of the Department of Electrical Engineering, then led by Professor Victor H. Rumsey. One area of effort involved the phase adjustment of the focal line array and the illumination of the reflector. Professor Yuen Tse Lo undertook the detailed design of the array, taking into account the appropriate illumination tapering to minimize sidelobe levels. The number of elements needed in a uniformly-spaced array 122 meters long (actually 137 meters, to compensate for foreshortening at maximum zenith angles) was daunting and the dynamic range of element currents required to provide proper illumination tapering promised to be difficult to achieve. Lo and I decided to try to achieve this tapering in part by use of nonuniform element spacing.
Design of the array element was done by Professor John D. Dyson. The Antenna Lab team were then engaged in their pioneering studies of frequency-independent antennas, and the conical log-spiral proved to be appropriate, having circular polarization with a low axial ratio over a wide angle from the axis of the cone. Its primary radiation pattern gave the proper tapering for illuminating a parabola with the chosen f/d ratio. Dyson's design gave a precise adjustment of phase with mechanical rotation--no difficulty with phase adjustment was ever experienced during the operation of the telescope.
Despite its excellent axial ratio, the conical log spiral still has a finite negative component of circular polarization. If the array is correctly phased to direct the positively polarized beam at a given zenith angle, there will also be a smaller negatively polarized beam at the negative of the given zenith angle. To eliminate this negative beam, Lo proposed an ingenious scheme. The feed lines to the several elements were of random lengths, and the phase for the positive polarization was corrected by mechanical rotation of the element. This merely doubled the random error for the negative polarization, and the negative beam never formed. These design efforts led Lo into his well-known studies of nonuniform and random arrays.
I also asked Lo to study the scattering of electromagnetic energy from simplified models of steel and of wooden towers to support the focal-line structure. His answer was that the two materials would have about the same net effect, so we decided to build the structure of wood, which we expected to be somewhat less expensive and to require less maintenance. After receiving the wooden structure design from the consulting engineers, we were dismayed to discover that each of the four towers contained four 155-foot-long, uniformly-spaced arrays of one-wavelength-long steel bolts. There wasn't time to calculate the cumulative effects of these hundreds of bolts, so we had them replaced by phenolic-impregnated wooden bolts. We also searched for suitable non-metallic guy ropes for the towers, but without success. We had to settle for steel guy cables and for steel elevator cables within one tower. In retrospect, these concerns were probably not important.
Feeding the array elements with the proper phases and amplitudes and with minimum attenuation proved to be a very difficult problem. Kwang-Shi Yang and I tried several feeding schemes. A full corporate structure feed system was considered too expensive by far and not easily realized because of the nonuniform spacing of the array elements. Instead, a traveling-wave transmission-line feed system was proposed, with the log-spiral elements loosely tapped into the line at the appropriate intervals, taking into consideration Lo's random-phase feeding system. This scheme proved to have insufficient bandwidth, so it was necessary to compromise, starting from the receiver port with a corporate structure, branching down to six ports feeding traveling-wave feed lines, each with several antenna elements. The individual elements were coupled to the transmission lines by adjustable capacitive probes. These latter were evolved through much experimentation by Yang and his assistants. Initially, inductive probes were used, but we were never able to make them work properly, so we abandoned them after making the first astronomical observations and redesigned the feeding hardware, much to the dismay of the astronomers who thought at the time that we were more interested in electronic experiments than in astronomy. Precise monitoring of the phase and amplitude at each element of the focal-line array was essential. The component of phase contributed by the individual log-spiral antenna element was precisely related to its mechanical rotation so it was only necessary to determine the phase and amplitude at the antenna port. Adapting a scheme he and his colleague Govind Swarup had developed for the Stanford Microwave Radioheliograph, Yang substituted an audio-modulated diode for the antenna at a given feed-system port and sent a signal into the line from the central receiver port. The phase delay of the audiomodulated reflection from the individual antenna port could then be compared with the phase of the signal generator. Adjustment for the proper co-phased condition for the broadside (vertical) beam direction could be made by rotating the index of the dial plate on the mechanical axis of the antenna element. Adjustment for an arbitrary zenith angle is made by further rotation with respect to this index, according to a table computed by Lo using the original Illiac computer. Design of the array transmission-line system probably took more engineering effort than any other task of the whole program. Every detail of the system had to be designed from scratch, as commercial components generally were unsuitable. In addition to the conical log-spiral antennas, these components included the coaxial transmission lines and all their hardware, including matching stubs, insulators, couplers, terminations, reference loads, refrigeration systems for reference loads, etc.
An additional task of comparable magnitude was the design and construction of the receivers and recording system. This also was a major task of K. S. Yang's, assisted by Kenneth Seib and several technicians. The Zenith Radio Corporation contributed electron-beam parametric amplifiers through the good offices of Dr. Robert Adler. These state-of-the-art amplifiers served well during the early years of observing, but they were difficult to maintain and adjust and they were eventually replaced by field-effect transistors when the latter became available. Maintaining proper balance between the effective temperatures of reference noise sources and the ambient-temperature-dependent losses in the hundreds of feet of transmission line in the feed system required many months of tinkering and adjustment.
The first installment of the radio source catalog was the PhD thesis of John M. MacLeod, published in 1964. Subsequently, most of the accessible parts of the sky (within 30 degrees of the zenith) were catalogued under the direction of Professor John R. Dickel, who also mapped a number of galactic extended sources, and by Professor John C. Webber. Professor Harold D. Webb also mapped considerable portions of the Milky Way galaxy, so that most of the accessible sky was covered at least once during the years 1959-1969. Among the many new cosmic radio sources discovered in this survey was the source VRO 42.22.01, which was later identified with the optical object BL Lacertae. This is the very small diameter core of an extended galaxy, and is the prototype of a large class of such objects whose extraordinary energetics are still unexplained. Two previously unknown supernova remnants were also discovered, expanding shells of gas expelled by exploding stars. A detailed study of the astrophysically interesting Cygnus X region of the Milky Way was begun, which has since been extended to other frequencies on other instruments. Many galactic H II (ionized hydrogen) regions were mapped.
One of the features of the original design concept was the provision of multiple reception beams, which would have permitted much more rapid surveying of the sky. The parabolic cylindrical antenna is inherently suited to multi-beam operation. Outputs of various sections of the focal-line array can be combined in several different beam-forming networks with appropriate phase shifts to produce a different beam angle for each network. The 400-foot telescope was ideally suited to this mode of operation: as there were six sections to the focal line array six closely-spaced beams were simultaneously possible, so that surveying could have proceeded at six times the rate of a single beam. It was never possible to secure funds to provide additional phase shifters and receivers (two per beam), so the catalog had to proceed at the most deliberate pace.
The reflector surface was of earth, precisely figured to 1/20th wavelength, covered with a heavy tarpaper liner topped with a galvanized steel one-inch (l/20) mesh. The earth base was subject to weathering. A period of each summer was reserved for precise surveying of the surface and for any necessary repairs which were effected with shovels, rakes, hoes and a portable tar kettle. This work and all other mechanical maintenance were supervised by Arno H. Schriefer. About $10,000 per year was required for mechanical maintenance, about two percent of the capital investment. Eventually, however, at about the time in 1969 that the sky coverage was being completed, there had accumulated substantial net erosion of earth from the upper slopes into the vertex of the parabola. This process resulted in progressive increase in the focal length, which eventually exceeded the available range of adjustment in the height of the focal line array. At that time the instrument was abandoned, its mission essentially complete.
In retrospect, several things probably should have been done differently. The variable-spacing tapering of the focal-line array saved many array elements, speeding both initial and daily phase adjustments and easing the amplitude tapering problem. However, it inevitably produced higher sidelobes in the meridian plane, which necessitated much re-observing of questionable sources. As the daily phase adjustment could never be precisely the same, the sidelobe levels differed from day to day, so re-observing usually answered any questions. Nonetheless, it probably would have saved time and money in the long run if a uniform array had been used. A more determined effort should have been made to institute a multi-beam system, though it's by no means certain funds for this could have been secured as the program was chronically underfunded. Less effort should have been expended on keeping metal components out of the near field of the reflector; this was probably a needless concern. More effort should have been made to acquire a workable digital data recording system, despite the nascent state of that art.
Other technical aspects of the adjustment were very successful. The phase adjustment system worked well and there was never any problem with beam pointing. The instrument was mechanically stable and unaffected by weather. It observed through snow, ice and rain storms and through a substantial earthquake without any visible modulation of the data.
The average starting salary for students graduating with a bachelor's degree in computer engineering (2016-17).